2020 – in progress
Ethical and Legal issues of Mobile Health‐Data – Improving understanding and eXPlainability of digitaL transformAtion and data technologies using artificial IntelligeNce
Motivation
This project explores the highly important topic of mobile data (mData) use in medicine and the employment of machine learning (ML) and deep neural networks (DNN) in this context. Clinical, research or other secondary uses involving mobile health data or ML, and even more the combination of mData and ML, have raised a panoply of new concerns creating legal and ethical barriers which interfere with trust of patients and society. Data protection concerns increase disproportionally with mData. ML could also be incompatible with the EU General Data Protection Regulation (GDPR) that postulates a right to explanation while ML algorithms function like a “black box”.
These concerns significantly impede advances in research: health care institutions collect high volumes of very useful data, including mData, but beneficial analysis remains scarce. While appropriate infrastructures are currently being developed, there is an urgent need for (i) clarification of the new pressing questions related to the ethico-legal governance of mData and the use of ML and (ii) an in depth exploration of patient concerns.
The topics of mData and ML are intrinsically related as the volume of mData collected during health care is exploding. Examples are smartwatch apps used in cardiology and non-invasive home-ventilators (continuous positive airway pressure, CPAP) collecting a variety of sleep, breathing and activity related parameters. AI such as ML is a promising tool to analyze this high data volume for research or clinical purposes.
While the global importance of AI is widely recognized for analysis of digitalized data in the clinical and/or research context – e.g. scans, skin lesions, ECGs, vital signs – challenges such as bias, data protection, and lack of transparency and explainability raise concern. Data access is debated, as manufacturers and health insurers are highly interested in mData. These unresolved issues interfere with trust and the efficient and beneficial implementation of these promising technologies.
Ethico-legal issues, including risks and benefits related to ML, vary enormously between different examples. Therefore, it is important to study them in concrete applications.
Our approach
This project will fill an important gap through a pragmatic approach on mData research and ML, by examining these questions related to 2 existing types of data collected in Swiss hospitals. The first are mData collected via smartwatches. The second type consists of perioperative data collected via patient data management systems (PDMS) at Swiss University hospitals. Both datasets are producing high amounts of interoperable data (ECG, vital signs etc.) where ML is highly useful. The 2 datasets will be compared to examine which additional issues exist for the use of mData as compared to clinical and secondary uses for more classical hospital data.
Collaboration
Prof. Dr. med. Bernice Simone Elger University of Basel, Switzerland.
Prof. Dr. med. Luzius Steiner University Hospital Basel, Switzerland.
Prof. Dr. med. Jens Eckstein, PhD University Hospital Basel, Switzerland.
Funding
SNF – Swiss National Science Foundation: National Research Programme NRP77